First records of Sturnira bakeri Velazco & Patterson, 2014 (Chiroptera: Phyllostomidae) from Colombia

We evaluate the occurrence of S. bakeri in Colombia, a recently described species. We report seven new records and include data on skull measurements of these individuals and information on the new localities. A discriminant analysis suggests that condyloincisive length and dentary length are the most important measurements to separate S. bakeri and S. luisi from S. lilium . However, to distinguish S. bakeri from S. luisi , we used discrete characters proposed in the original descriptions of these two taxa. Sturnira bakeri should no longer be considered a regionally endemic species from Ecuador and Peru.

Sturnira Gray, 1842 is a genus of frugivorous bats distributed across the Neotropics from Mexico to northern Argentina, including some islands in the Caribbean (Velazco & Patterson 2013). Species within this genus are important during successional stages in Neotropical forests (Muscarella & Fleming 2007), establishing mutualistic associations mainly with plants of the genus Piper and Solanum (Saldaña-Vázquez et al. 2013;Montoya-Bustamante et al. 2016).
With 23 recognized species, Sturnira is the most species rich genus in the family Phyllostomidae (Velazco & Patterson 2014;Molinari et al. 2017), which had its initial radiation in the Andes of South America (Velazco & Patterson 2013). As a product of a recent taxonomic revision of the genus, two new species were described: S. bakeri and S. burtonlimi (Velazco & Patterson 2013. Currently, both of these taxa are represented by a limited number of specimens; thus, S. bakeri was thought to be endemic to southwestern Ecuador (Velazco & Patterson 2014;Tirira 2015).
The type series of S. bakeri includes specimens from two localities in southwestern Ecuador: Palmales, Reserva Militar Arenillas, El Oro (03°40ʹ27.4ʺ S, 080°06ʹ20.0ʺ W; elevation: 49 m) and Quebrada Seca, Fuerte Militar Arenillas, El Oro (03°39ʹ24.1ʺ S, 080°10ʹ56.2ʺ W, elevation: 43 m). Therefore, there was a high probability that this species inhabits northwestern Peru rather than elsewhere (Velazco & Patterson 2014). Recently, Sánchez & Pacheco (2016) tested this hypothesis and reported the presence of S. bakeri from six localities in northwestern Peru. No further information has been published on S. bakeri and its geographic distribution remains obscure. Our aim was to evaluate the occurrence of S. bakeri in Colombia, as well as to increase our knowledge of the morphological variation and distributional pattern of this species.
An exhaustive review of Sturnira specimens deposited in the mammal collection of Universidad del Valle (UV) was conducted using the original description of S. bakeri (Velazco & Paterson 2014). We compared putative specimens of S. bakeri with representatives of morphologically and phylogenetically closely related species in the genus (Sturnira luisi Davis, 1980 andSturnira lilium (É. Geoffroy, 1810)). Studied specimens are listed in the Appendix. Although Velazco & Patterson (2013) restricted the distribution of S. lilium to the Brazilian Shield, they did not clarify its status in Colombia (Ramírez-Chaves et al. 2016). Therefore, in this study we still identify as S. lilium those specimens that agree with its description.
Date, sex and information on capture localities were taken from museum tags. Whenever specimen coordinates were not available in museum labels, they were calculated through direct projection using Arcgis 10.2.2 software. Georeferencing was made by association to a physical element (single point), and each locality was located in the map (Table 1; Figure 1). Cartography from SIGOT's data bank was used, and georeferencing was based on MAGNA SIRGAS-WGS84 systems.
Standard mammal measurements were taken from each museum specimen's tag, and if not available were measured directly when possible (Table 2). Cranial measurements were taken from S. bakeri (n = 7), S. lilium (n = 19), and S. luisi (n = 18) using a Mitutoyo digital caliper (to the nearest 0.05 mm) following Velazco & Paterson (2014 bakeri (Velazco & Paterson 2014), a linear discriminant analysis was performed to assess morphometric differences between S. bakeri and its closely related species. Total length, hindfoot length, ear length and weight were not included in the analysis because they were not available for all specimens. Mahalanobis distances and its posterior classification probabilities were used to determine if individuals were correctly assigned to each particular group. We document the presence of seven Colombian specimens representing S. bakeri. These specimens matched the morphological description of the species and were previously misidentified as S. luisi (n = 6) and S. lilium (n = 1) ( Table 1). One of them (UV 10817), despite having all the other skull and skin traits characteristic of S. bakeri, does not have bicuspidate I1 and there is no sign of wear. Notwithstanding, it was collected at the same date and place of UV 10818, a typical S. bakeri ( Figure 1). Dorsal fur varied from dark brown to a light yellowish brown, while ventral fur varied from pale brown to a yellowish beige. The presence of yellowish epaulettes is conspicuous in all specimens, males and females, with the exception of specimen UV 10818.
Comparative measurements and weights from the holotype of S. bakeri (Velazco & Patterson 2014) and S. bakeri individuals reported herein are shown in Table 2. The linear discriminant analysis revealed morphological differences among the three analyzed groups (Figure 2). The first root explained 81.8% of the variability among groups, separating S. bakeri and S. luisi from S. lilium. The second discriminant function explained 18.2% of the variability, allowing us to differentiate S. bakeri from S. luisi. Condyloincisive length (CIL) and dentary length (DL) were the most important measurements to separate S. lilium from S. bakeri and S. luisi (standard coefficients for root 1: CIL = -1.40, DL = -0.99). Condylocanine length (CCL) and dentary length (DL) were the most important measurements to differentiate S. bakeri from S. luisi (standard coefficients for root 2: CCL=2.23, DL = -1.84). The posterior probabilities classified all the studied individuals according to their a priori assignment.
Here we present the first records of S. bakeri from Colombia. Six new localities were identified for this taxon in Colombia, extending this species' range by 1,034 km north from previous closest record (Table 1). Altitude varied from 400 to 2,000 m, and three life zones were identified for the species distribution: premontane wet forest, premontane rain forest, and lower montane wet forest.
According to Velazco & Patterson (2014), S. bakeri might be confused with S. lilium and S. luisi, but our morphologic analyses show it is easily recognizable by its skull. The most useful characters we used to distinguish S. bakeri from S. lilium and S. luisi were: a globular braincase with a slender rostrum (Figures 3, 4); a well-developed sagittal crest ( Figure 5); an oval sphenorbital fissure ( Figure 6); an absent anterior process of the glenoid fossa ( Figure  7); well-developed clinoid processes ( Figure 8); usually bicuspidate upper inner incisors with a small lateral cusp ( Figure 9); well-defined metaconid and entoconid in m1 and m2, which are separated by a deep notch (Figures 10,  11); and tricuspidate lower incisors; all of these characteristics were proposed by Velazco & Patterson (2014). We did not find these traits to be variable, as mentioned by Sánchez & Pacheco (2016).
Despite the upper inner incisors and their cusps have been suggested as useful characters to differentiate other species of Sturnira (e.g., S. oporaphilum and S. ludovici from S. hondurensis and S. burtonlimi; Velazco & Paterson 2014), they seem to be variable in our series of S. bakeri. Although, upper inner incisors of specimen UV 10817 did not match the description of S. bakeri, its classification  Table 2. Comparative measurements (mm) and weights (g) of the recorded Sturnira bakeri from Colombia and the type series. Measurements were taken according to Velazco and Patterson (2014). GLS: greatest length of skull, CIL: condyloincisive length, CCL: condylocanine length, BB: braincase breadth, ZB: zygomatic breadth, PB: postorbital breadth, MB: mastoid breadth, MTL: maxillary toothrow length, WM2: width at M2, DL: dentary length, MDTL: mandibular toothrow length, FL: forearm length, TL: total length, HL: hindfoot length, EL: ear length, W: weight.   within this taxon was supported by the other morphologic characters mentioned above, and by our discriminant analysis as well. Velazco & Patterson (2014) also proposed some external characters useful to identify S. bakeri, including hair color and length. However, as noted by Sánchez & Pacheco (2016) pelage traits may be very variable within this species, as in other Sturnira species (Tamsitt et al. 1986). Thus, skin characters should be used only to complement identifications based on skull characters and should not be considered as diagnostic.

Measurements
Results from the discriminant analysis show that S. bakeri and S. luisi can be distinguished from S. lilium using condyloincisive length and dentary length, which are smaller in S. lilium (Table 3). Despite condylocanine length and dentary length were the most important variables to differentiate S. bakeri from S. luisi, these measurements overlap ( Table 3), suggesting that these variables should not be the only ones used to separate these species. However, the correct classification of groups using the posterior probabilities suggest that the skull measurements and traits proposed by Velazco & Patterson (2014) (with the exception of bicuspidate I1) are the best way to discriminate S. bakeri from S. luisi.
Our findings confirm the presence of S. bakeri in Colombia, and add new information on its distribution and morphological variation within this taxon. Contrary to the type locality, our specimens were collected in wet and rain forests. In Colombia, lower montane wet forests are widely distributed across the Andes; their mean temperature varies from 12-18 °C, and mean annual rainfall oscillates from 2,000-4,000 mm (IGAC 1988).
On the other hand, mean temperature at premontane wet forests varies from 18-24 °C, and mean annual rainfall varies from 2,000-4,000 mm; in Colombia, this life zone can be found especially in the coffee zone. At last, premontane rain forests are principally located in the eastern slope of Eastern Andes cordillera and the western slope of Western Andes cordillera (Chocó Biogeográfico) (IGAC 1988). The latter, the Chocó Biogeográfico, is considered as a "hotspot" for phyllostomid bats (Mantilla-Meluk et al. 2009), and is one of the most humid ecosystems in the world due to interception of coastal winds (Kattan et al. 2004). Its mean temperature varies from 18-24 °C, and mean annual rainfall varies from 4,000-8,000 mm (IGAC 1988).
Colombian localities of S. bakeri, specifically those from the Chocó Biogeográfico (UV 2150, 4136, 4540, 10817 and 10818) are characterized by high plant diversity with a good representation of genera included in Sturnira diet, such as: Vismia, Piper, and Cecropia (Faber-Langendoen & Gentry 1991;Rangel-Ch. et al. 2011;Lobova et al. 2009;Montoya-Bustamante et al. 2016). Specimen UV 11932 was collected in a highly disturbed locality, characterized by the presence of coffee and banana crops with a riparian forest relict, near Bugalagrande river (V. Rojas-Díaz com. pers.). This new information suggests that S. bakeri in Colombia occupies a wide variety of habitats, including highly disturbed areas by human activities. These new localities for S. bakeri increase substantially the known distribution of this taxon that should no longer be considered as endemic.
Finally, our study represents an important example of the significance of collecting specimens for biological collections and of the relevance of collection-based studies, a matter recently in debate (Minteer et al. 2014;Rocha et al. 2014). Although S. bakeri was recently described, specimens in the UV collection had been deposited there more than 30 years ago and prove that biological collections are an invaluable and continuous source of new information on biodiversity and natural history.
In conclusion, our findings add to the knowledge of the ecology of S. bakeri, previously only known from dry forests. Our data support the occurrence of this species in premontane wet forests, premontane rain forests, and lower montane wet forests. The distribution of S. bakeri is not as restricted as originally thought. Further studies should clarify the nature of the distribution of this species and whether it has a continuous or discontinuous distribution. With these new records the known mammalian species richness from Colombia increases to 521 (Solari et al. 2013